Yummy Dream

Inspiration

Dreams are full of creativity and imagination, but they’re often fleeting. We wanted to transform those abstract, playful ideas into something tangible and fun. The idea came from wondering: “What if AI could turn dreams into recipes?” This project combines creativity, AI, and interactivity to bring dreams to life in a unique way.

What it does

The Dream-to-Recipe Generator lets users type a dream. Our AI then generates a complete, imaginative recipe including:

• Recipe Name inspired by the dream
  
• Ingredients with quantities
  
• Step-by-step instructions
  

How we built it

• AI Model: GPT-OSS-20B to generate recipes from dream descriptions.
  
• Prompt Design: Custom prompt templates ensure consistent output formatting.
  

Challenges we ran into

• Model Size & GPU Limits: GPT-OSS-20B is very large, so running it in Colab required GPU acceleration and careful memory management.
  
• Output Consistency: Ensuring the AI consistently produced readable recipes with proper ingredients and steps.
  

Accomplishments that we're proud of

• Created a system that turns abstract dreams into tangible, creative recipes.
  
• Developed an effective prompting strategy to keep AI outputs fun, structured, and readable.
  

What we learned

• Hands-on experience with large language models and prompt engineering.
  
• Practical lessons on GPU memory management and handling large model outputs.
  

What's next for Yummy Dream

• Add customization options like dietary preferences or cuisine types.
  
• Deploy as a standalone web app accessible without Colab.
  
• Include AI-generated visuals or illustrations alongside recipes for a richer experience.
  

Built With

• google-colab
• python